Program Director for Biomedical Imaging Informatics

June 12, 2012

Vinay Pai, Ph.D.

National Institute of Biomedical Imaging and Bioengineering

National Institutes of Health

Democracy II, 6707 Democracy Blvd, Ste. 200

Bethesda, MD 20817

RE: Supplemental Response: National Alliance for Medical Image Computing (NA-MIC) Annual Progress Report

Dear Vinay,

I have discussed with the NA-MIC team the questions raised by the reviewers, and have provided our responses in the attached document. We have interwoven the queries with our responses.

Please let me know if you have additional questions.

Sincerely,

Ron Kikinis, MD, PI

NA-MIC RESPONSE TO QUERIES

We appreciate the careful review of our progress report and have included our responses below. We regret not having provided this level of detail in the original progress report, especially since it clearly contributed to a wide gap between our own and the reviewers’ assessment of the productivity of our DBPs. In the 8-year history of NA-MIC, this is our third set of DBPs, and we are particularly proud of the level of results they have achieved within their first two years – as measured by their publications as well as their contributions to the open source software platform 3D Slicer – even when compared to their predecessors, who were all strong contributors in their own right. Internally, our team sentiment is that working with two previous cycles of DBPs greatly helped us refine our processes and contributed to the rapid progress made by the current DBPs, which is why the reviews were particularly unexpected. In the rest of this document, we have provided responses to each of the points raised by the reviewers, as well as a set of milestones for the upcoming year for each of the DBPs.

A.Huntington's Disease, Hans Johnson, PI, University of Iowa

a1.Do the researcher's think that these tools will be applicable in other domains, for example, neurodevelopmental studies in autism or schizophrenia? This might be a very useful tool for these mental diseases.

Response: The tools developed for the HD-DBP already are having a significant impact on a wide range of other neurodevelopmental and neurodegenerative studies. The University of Utah (Utah) and University of North Carolina (UNC) NA-MIC algorithmic work concerns the development of novel image processing and analysis tools that center on 4D longitudinal analysis (Huntington's) and automated segmentation of complex pathology in single and multi-time-point patient data. These computational tools are generic and can be adapted to a wide range of clinical applications.

At the University of Iowa, many of the tools are being actively used to investigate white matter changes in cleft lip and palate, atherosclerotic vascular disease, anorexia nervosa, juvenile HD, cognitive impairment in aging, and schizophrenia. UNC has applied the DTI framework, still under development in NA-MIC for Huntington's Disease, to studies of drug abuse (a. cigarette smoking; b. chronic alcoholics). The Utah group is using these tools and software to investigate the gene-brain phenotype relationship in Down's Syndrome, where studying brain growth and growth alterations due to genetic differences are the key research issues. A joint effort by the UNC/Utah groups has applied these tools to study infant brain development related to autism, with the objective of comparing subjects at high- versus low-risk of developing autism (from the Autism Centers of Excellence/ACE project).

The cross-fertilization of NA-MIC technology with the aforementioned grants is best reflected in the most recent joint publications of these groups with NA-MIC partners. This list, which is extracted from a much longer history of collaborative research, includes only the most recent works.

Published Articles

Geng X, Gouttard S, Sharma A, Gu H, Styner M, Lin W, Gerig W, Gilmore JH. Quantitative Tract-Based White Matter Development from Birth to Age Two Years. NeuroImage, pp. 1–44,Mar. 2012.

Wolff JJ, Gu H, Gerig G, Elison JT, Styner M, Gouttard S, Botteron KN, Dager SR, Dawson G, Estes AM, Evans AC, Hazlett HC, Kostopoulos P, McKinstry RC, Paterson SJ, Schultz RT, Zwaigenbaum L. Piven L. Differences in white matter fiber tract development present from 6 to 24 months in infants with autism. Am J Psychiatry 2012; Feb 17; Epub ahead of Print. 10.1176/appi.ajp.2011.11091447.

Fishbaugh J, Prastawa M, Durrleman S, Piven J, for the IBIS Network and Gerig G. Analysis of Longitudinal ShapeVariability via Subject Specific Growth Modeling, accepted MICCAI'12 conference, in print Springer LNCS, to appear Oct. 2012.

Neda Sadeghi, Marcel Prastawa, P. Thomas Fletcher, John H. Gilmore, Weili Lin, Guido Gerig, 'Statistical growth modeling of longitudinal DT-MRI for regional characterization of early brain development.' Proc. IEEE ISBI 2012.

Sharma A, Durrleman S, Gilmore JH, Gerig G. 'Longitudinal Growth Modeling of Discrete-Time Functions With Application To DTI Tract Evolution In Early Neurodevelopment.' Proc. IEEE ISBI 2012, to appear May 2012, in print.

Timpe JC, Rowe KC, Matsui JK, Magnotta VA., Denburg NL. (in press). White matter integrity, as measured by diffusion tensor imaging, distinguishes between impaired and unimpaired older decision-makers. Journal of Cognitive Psychology, in press.

Articles in preparation

Rowe KC, Matsui JT, Mayberg HS, Magnotta VA, Arndt S, Johnson HJ, Nopoulos P, Paradiso S, McCormick L, Fiedorowicz JG, Epping EA, Moser DJ. (in preparation). Depressive symptoms related to low fractional anisotropy in the right ventral anterior cingulate in vascular disease. American Journal of Geriatric Psychiatry.

Rowe KC, Arndt S, Magnotta VA, Nopoulos P, Paradiso S, Matsui JT, Johnson HJ, Moser DJ. (in preparation). Characterizing white matter health in atherosclerotic vascular disease. Stroke.

Rowe KC, Stillman AS, Arndt S, Fiedorowicz JG, Haynes W, Moser DJ. (in preparation). Diffusion tensor imaging correlates to anxiety symptoms in atherosclerotic vascular disease. International Journal of Geriatric Psychiatry.

Abstracts using neuroimaging tools

Moser DJ, Rowe KC, Magnotta VA, Haynes W, Nopoulos P. Cerebral White Matter Volume is Associated with Forearm Vascular Function in Elderly Men with Vascular Disease. American College of Neuro- Psychopharmacology Conference, 2010.

Timpe J, Rowe KC, Matsui JT, Magnotta VA, Denburg N.L. White matter integrity as measured by diffusion tensor imaging distinguishes between impaired and unimpaired older adult decision-makers. University of Iowa Medical Student Research day, 2010.

Rowe KC, Vitense K, Axelson E, Magnotta VA, Nopoulos P, Moser DJ. Diffusion Tensor Imaging: revealing hidden connections between vascular disease, emotion, and cognition. University of Iowa Neuroscience Research Week, 2011.

Rowe KC, Magnotta VA, Matsui JT, Vitense K, Axelson E, Brumm M, Arndt S, Paradiso S, Nopoulos P, Moser DJ. Diffusion tensor imaging in the atherosclerotic brain: implications for depression, processing speed, and attention. International Society of Vascular Behavioral and Cognitive Disorders Conference, 2011; Society for Neuroscience Conference, 2011.

Rowe KC, Magnotta VA, Matsui JT, Vitense K, Axelson E, Brumm M, Arndt S, Paradiso S0, Nopoulos P, Moser DJ. Relationships among white matter health, depression, and cognition: A diffusion tensor imaging study. Cognitive Neuroscience Society Conference, 2012.

a2.If yes to question a1, what would be the requirements for training neuroscientists in these fields to use the tools?

Response: We already have held workshops for the use of Slicer for autism and schizophrenia, both of which were our previous DBPs. We picked HD for the current DBP to round out our tools for neurodevelopmental disorders. The teams at Utah and UNC that participated in the previous DBPs continue to work with the HD group to ensure that the relevant tools continue to benefit autism and schizophrenia.

To reiterate, training is a recognized and integral part of the HD-DBP effort, and we routinely customize our approach to accommodate a broad range of individuals with diverse educational backgrounds. For training neuroscientists working in fields other than HD, we only require that they have basic familiarity with using computer software and graphical user interfaces. Using this as a foundation, we are confident that we can train them to use Slicer software by participating in interactive Slicer workshops, where they directly participate using their laptop computers and any sample data of interest.

This has been a successful model, as indicated by the long-standing record of the NA-MIC PI (Ron Kikinis) and several senior investigators in collaborating with neuroscientists to provide working computational solutions for analyzing different neurological disorders.

Other senior investigators, such as Guido Gerig, Martin Styner and NA-MIC Training PI, Sonia Pujol, also have extensive experience in training scientists in the algorithmic aspect of the computational tools by conducting workshops, which allow Slicer users to develop intuition of the underlying theories so they can analyze their data effectively.

To address the requirements from neuroscientists and other clinical researchers, we continuously interact with clinical collaborators. Our senior investigators have distilled from these interactions a set of criteria that drive all tool development in NA-MIC. These criteria include creation of intuitive user interfaces, documentation, sample datasets and processing protocols, examples and guides, prototyping tools / sandbox environments, etc. Once tools are developed using these criteria, we then launch into training-specific tasks that include creating good documentation, tutorials, focused workshops, and user-friendly toolsets like SimpleITK. Below we list some specific training activities.

For example, in addition to the NA-MIC Project Events held each year, the HD-DBP team has given presentations at the MICCAI conference in Toronto, Canada, with approximately 40 participants. We also conducted an intensive 2-day workshop ( at our institution (University of Iowa), with 35 participants from 6 different departments (psychiatry, radiology, electrical and computer engineering, biomedical engineering, neurology, radiation oncology). The success of that training has prompted the demand for another similar session to be held August 8-9, 2012. A series of 5 lectures were given to the Iowa Institute of Biomedical Imaging highlighting the NA-MIC tools and describing how to access and use them.

Guido Gerig is organizing a workshop for spatio-temporal image analysis (STIA'12, http://www.sci.utah.edu/stia2012-home.html). NA-MIC activities to further advance theory and methodology for 4D image processing will be the key issues. At this workshop, we will emphasize the NA-MIC efforts as a key example of collaborative algorithmic, engineering, and driving clinical work. Also, we will feature the HD and TBI projects as excellent examples to drive future research, and we will point the audience to NA-MIC resources including software and test datasets.

NA-MIC Training PI, Sonia Pujol, is working with the NA-MIC community to organize the DTI challenge (with input from Ron Kikinis, Guido Gerig, Martin Styner and others).

(http://projects.iq.harvard.edu/dti_challenge/pages/data).

Our work integrating SimpleITK and Nipype into our methodologies allows neuroscientists easy access to the algorithms available in the NLM Insight segmentation and registration toolkit (ITK). Integrating the HD-DBP tools with these frameworks has two distinct advantages: (1) they provide a level of documentation that is easy and familiar to a large number of neuroscientists, (2) they provide a convenient distribution mechanism for others to access and deploy.

For example, we are working to make the morphometric analysis pipeline available as part of the Nipype distribution. This pipeline is complex and is composed of various registration, resampling, and segmentation steps. The integration of this complex pipeline, and any Slicer complaint modules, is a streamlined process made possible by an automated integration engine developed by the NA-MIC team.

b.In Table 1 (XNAT HD Data Collaborators), a number of the highlighted scientists work in neuroscience applications. Are they aware of the longitudinal capabilities being developed?

Response: Yes, we are working closely with several of these teams to direct their efforts to maximize the integration with existing HD-DBP efforts. The efforts of Thomas Schutz (Max Planck Institute for Intelligent Systems, Tubingen, Germany) has resulted in a MICCAI 2012 paper, and more importantly, an improved DWI tool for “Learning a Reliable Estimate of the Number of Fiber Directions in Diffusion MRI.” We continue to collaborate with him in the hopes of using that tool on the larger HD datasets. We are working with Polina Golland (MIT) to generate preliminary data needed to apply for the funding necessary to accomplish a connectivity analysis of DWI and resting state data. A collaboration between Gary Christensen (University of Iowa) and Anuj Srivastava (Florida State University) and the HD-DBP is investigating the development of advanced shape statistics for longitudinal analysis.

In addition to the non-HD neurodevelopmental studies, the HD-DBP is having a significant impact on the larger international HD community. In particular, the international TRACKON study of HD center at the University College of London has been aggressively following and adopting NA-MIC tools and NA-MIC best practices into their prospective study. Additionally, a separate Juvenile HD project is adopting the tools developed by the HD-DBP as well.

Articles of external HD community (beyond the scope of the HD-DBP)

Aylward EH, Nopoulos PC, Ross CA, Langbehn DR, Pierson RK, Mills JA, Johnson HJ. (The U. of I., et al. (2011). Longitudinal change in regional brain volumes in prodromal Huntington disease. Journal of neurology, neurosurgery, and psychiatry, 82(4), 405-10. BMJ Publishing Group Ltd. doi:10.1136/jnnp.2010.208264.

Scahill RI, Hobbs NZ, Say MJ, Bechtel N, Henley SMD, Hyare H, Langbehn DR, et al. (2011). Clinical impairment in premanifest and early Huntington’s disease is associated with regionally specific atrophy. Human brain mapping, 000(July 2010), n/a-n/a. Wiley Subscription Services, Inc., A Wiley Company. doi:10.1002/hbm.21449.

Tabrizi SJ, Reilmann R, Roos RAC, Durr A, Leavitt B, Owen G, Jones R, et al. (2012). Potential endpoints for clinical trials in premanifest and early Huntington’s disease in the TRACK-HD study: analysis of 24 month observational data. Lancet neurology, 11(1), 42-53. Elsevier. doi:10.1016/S1474-4422(11)70263-0.

Tabrizi SJ, Scahill RI, Durr A, Roos RAC, Leavitt BR, Jones R, Landwehrmeyer GB, et al. (2011). Biological and clinical changes in premanifest and early stage Huntington’s disease in the TRACK-HD study: the 12-month longitudinal analysis. Lancet Neurol, 10(1), 31-42. Elsevier. doi:10.1016/S1474-4422(10)70276-3.

Leveraging the HD-DBP to foster new research

Two more grants with PI John H. Gilmore on early infant brain development and risk factors for mental illness are making use of the newly developed tools for spatio-temporal structural and DTI analysis. A renewal application for the Utah Twin grant is in preparation.

Two grants w.r.t. ancillary studies of Hungtington Disease Image Data (PAR-12-097), centered about NA-MIC image analysis technology on 4D shape analysis (PI Utah G. Gerig) of anatomical structures on 4D DTI analysis (PI UNC M. Styner), were submitted in April 2012 and are currently in review.

1 R01 HD055741 (Autism Centers of Excellence ACE, project IBIS) – PI is Joseph Piven (still in progress, awaiting final decision on renewal)

1 R01 HD067731 (Down's syndrome) -- PI is Julie R. Korenberg, Utah (awarded September 2011)

2 P50 MH064065-06 -- PI John H. Gilmore, Prospective Studies of the Pathogenesis of Schizophrenia, Silvio O. Conte Center for the Neuroscience of Mental Disorders

R01 MH070890 -- PI John H. Gilmore, Prospective studies of Early Brain Development in Twins (renewal in preparation)

c.Could SPHARM be used in longitudinal studies in autism and schizophrenia?

Response: Yes, absolutely. The methodology developed within core 1a in collaboration with the Huntington's DBP is generally applicable and thus is not specific to Huntington's disease. As an example, we submitted a paper on applying the longitudinal SPHARM analysis on normal brain development (see attached paper, submitted to MICCAI workshop on perinatal and postnatal imaging). Its application to autism in two studies (a. former NA-MIC DBP with a longitudinal study from 2-4 years of age, b. Autism Center of Excellence network on brain development in high-risk subjects from 6 to 24 months) at UNC is ongoing.

Additionally, the Utah group is developing an alternative shape analysis framework that addresses some limitations in SPHARM, where analysis is conducted without correspondence and using low dimensional parameters. This new framework enables the computation of average shapes as well as the deviations from the average (variability) in a mathematically sound and convenient manner. They are applying this to HD and have preliminary results (see Figure 2). Preliminary results have also been generated for comparing the deep brain structures of children with Down's syndrome and controls, and this will appear in an upcoming MICCAI paper:

Durrleman S, Prastawa M, Korenberg JR, Joshi S, Trouvé A, Gerig G. Topology Preserving Atlas Construction from Shape Data without Correspondence using Sparse Parameters. Medical Image Computing and Computer Assisted Intervention (MICCAI) 2012, in print Springer LNCS, to appear Oct. 2012.

Figure 2. Intra-subject subcortical shape changes.

B.Adaptive Radiotherapy for Head and Neck Cancer, Greg C. Sharp, PI, MGH

a.The science officers were disappointed by the minimal progress shown with this DBP. A detailed discussion on this is attached (quoted here).

"The problems addressed in this project regarding deformable registration applied to proton treatments of H & N cancer are well established and previously investigated in the field by others using both vendor produced products and academically developed ones [ see 1) ‐ 3) below]. Thus the significance lies more in the extension of the NA‐MIC / 3D Slicer software tools and core expertise to the radiation therapy field. Such an extension may well offer some new and unique features for adaptive radiotherapy (biological modeling?, integration of radiological or molecular parameters into RT planning?) and additionally it may bridge the gaps between the RT community and the much broader imaging community through its shared ontologies, algorithms and software tools.

Toward that end the “plans” for the coming year (page 211) are of the greatest interest and represent the highest potential impact of this project. “Improved support for radiotherapy structure sets in 3D slicer” has been a major impediment to its implementation in RT and thus other tools (mostly from the computer treatment planning vendor community) have already matured within the RT community. Similarly, for the “improvement and documentation of dose review tools” and the “interactive segmentation tools”. The PI should explain further which aspects of these developments will be unique compared to what has already come into common usage in the RT arena and what the timeline is for those novel developments."

Response: An important goal of this DBP is to add infrastructural capabilities to Slicer that will enable RT research. For example, while support for DICOM RT is not a novel functionality, RT researchers cannot leverage algorithms developed for image analysis by the rest of the computational image processing community without having full support and implementation for DICOM RT.